AI vs Machine Learning vs Deep Learning

Artificial intelligence (AI) is the broad field of building computer systems that perform tasks associated with human intelligence. Machine learning (ML) is a subset of AI in which systems learn patterns from data instead of following hand-written rules. Deep learning is a subset of machine learning that uses neural networks with many layers. Think of them as nested circles: all deep learning is machine learning, and all machine learning is AI, but not the other way around.

The relationship at a glance

Picture three circles, one inside the other:

• The outer circle is AI: any technique that lets computers perform tasks associated with human intelligence.
• Inside it is machine learning: AI that learns from data.
• Inside that is deep learning: machine learning that uses many-layered neural networks.

What AI is

Artificial intelligence is the umbrella term. The OECD, in a definition updated in 2023, describes an AI system as a machine-based system that infers from the input it receives how to generate outputs such as "predictions, content, recommendations, or decisions."

AI does not require learning. Many earlier AI systems relied on rules written by people. DENDRAL, developed at Stanford in 1965, is recognized as the first expert system: a program that captured specialists' knowledge as rules. Search methods are another classic AI approach. The A* search algorithm, developed by researchers working on the Shakey robot at the Stanford Research Institute in the late 1960s, is still used for tasks such as finding driving directions.

The LATHIC Glossary has the short definition of artificial intelligence.

What machine learning is

Machine learning is the part of AI where systems learn from examples rather than following rules someone wrote for every case. A machine learning model is trained on data, finds patterns and uses them to make predictions about new data. IBM researcher Arthur Samuel is credited with coining the term, which appears in his 1959 paper on a checkers program that improved by playing.

Machine learning is now the dominant way AI is built. As IBM summarizes the relationship: "All machine learning is AI, but not all AI is machine learning."

For how machine learning works step by step, see What Is Machine Learning?. The short definition is in the LATHIC Glossary entry for machine learning.

What deep learning is

Deep learning is the branch of machine learning built on neural networks with many layers. Google's machine learning glossary defines a deep model as "a neural network containing more than one hidden layer." Each layer transforms the data a little further, which lets deep networks learn complex patterns directly from raw inputs such as pixels, sound waves or text.

Deep learning drove the major AI advances of the 2010s. In 2012, a deep neural network won the ImageNet image recognition challenge with a top-5 error rate of 15.3%, compared with 26.2% for the next-best team. Yoshua Bengio, Geoffrey Hinton and Yann LeCun later received the 2018 ACM Turing Award for work that made deep neural networks a critical part of computing.

For how neural networks work, see What Is a Neural Network?.

AI vs machine learning

The difference between AI and machine learning is scope. AI is the goal and the field; machine learning is one method for reaching that goal, and today the most widely used one.

In everyday conversation, people often say "AI" when they mean machine learning, because nearly all modern AI products use it. The distinction matters when you want to know how a system works: whether its behavior was written by people or learned from data. That affects what it depends on, how it can fail and how it can be improved.

Machine learning vs deep learning

Deep learning sits inside machine learning. The main differences are the type of model, how much data and computing power it needs, and what kinds of problems it handles best.

Neither is universally better. For many business problems with structured data, simpler machine learning methods are effective and easier to explain. Deep learning excels when the data is complex and plentiful.

Where generative AI and large language models fit

Two more terms often appear alongside these three:

• Large language models are deep learning models. IBM describes them as "a category of deep learning models trained on immense amounts of data," and most are built on the transformer neural network architecture.
• Generative AI describes what a system does, creating new content, rather than the technique it uses. Most modern generative AI, including chatbots and image generators, is built with deep learning.

So ChatGPT sits in all three circles at once: it is AI, it was built with machine learning, and specifically it uses deep learning. For more on this category, see What Is Generative AI?.

How the terms are used in practice

You will hear these terms used loosely, and knowing the typical patterns helps:

• "AI" is the most common label in products, news and marketing, regardless of the method underneath.
• "Machine learning" tends to appear in data science and engineering discussions about training models and making predictions.
• "AI/ML" is shorthand used in job titles, teams and vendor materials to cover both the field and its main method.
• "Deep learning" usually comes up when the conversation turns to neural networks, image and speech recognition, or large language models.

When someone says a product "uses AI," a useful follow-up question is: does it learn from data, and if so, what data?

Common misconceptions

"AI and machine learning mean the same thing." Machine learning is one part of AI. AI also includes approaches that do not learn from data.

"All AI learns and improves by itself." Rule-based systems do not learn. Even machine learning models usually stay fixed after training until developers retrain or update them.

"Deep learning is always the better choice." Training deep learning models from scratch typically needs large amounts of data and computing power. Simpler machine learning methods often work better for smaller, structured datasets.

"Neural networks work like the human brain." Neural networks were loosely inspired by the brain, but their units are simple mathematical functions, and they learn in very different ways.